ALMOND, D.G. and VALDERRAMA, W. Performance enhancement toolf for grining mills.

International Platinum Conference Platinum Adding Value,

The South African Institute of Mining and Metallurgy, 2004.

Performance enhancement tools for grinding mills

D.G. ALMOND* and W. VALDERRAMA
*FFE minerals USA Inc., Performance Enhancement Technology, Bethlehem, PA USA
Federico Santa Maria Technical University, Department of Materials Science, Valparaiso, Chile

Operation of large SAG mills requires a change in the conceptual approach to grinding processes.
The traditional energy consumption criteria are now enhanced with new criteria derived from the
mechanical dynamics of the mill load. This new approach demands instrumentation capable of
providing accurate information about the load movement and the ratio of balls and ore inside the
mill. This paper discusses two new performance enhancing tools available to meet these needs.

The need for new instruments

Economies of scale play a major role in making new mining
projects economically viable and this has led to increased
equipment sizes. Although this has been particularly
evident in the size of haulage trucks and shovels used in the
mining operation, this phenomena has also filtered into the
area of minerals processing and into milling in particular.
SAG and ball mills have grown in size and there are now
many operational mills in the range of 2840 feet in
diameter with installed powers ranging between 820 MW.
The very high capacity of these mills and the common use
of variable speed drives often result in operational
problems. These problems are associated with high-energy
impacts striking mill liners resulting in poor energy
efficiency, as well as the fracture of steel balls, accelerated
wear, and even liner breakage. This has given rise to the
need for a new range of instruments aimed at providing
information to the operator about the load composition and
movement within SAG mills.
The concepts for operating SAG mills have been
inherited almost entirely from conventional ball mills.
However, these mills operated under a constant ball load
and at a fixed speed, the slurry load and properties being the
main sources of concern. Under these relatively stable
conditions, theoretical approaches based on energy amounts
required for the grinding process were the rule. Bond
developed his Work Index concept as a general property of
the ore. Later, a set of parameters Si and Bij were used for
describing grinding kinetics. All these approaches neglect
the details of load movement and how energy is actually
used inside the mill. For a long time, it has been
commonplace to consider a basic process rule that semiautogenous mills grind by impact, whereas ball mills do so
by attrition. As a consequence, SAG mills have been
designed and operated to generate a large number of
impacts by the falling load, without paying much attention
to the place where these impacts are occurring. A second
consequence has been that as long as the mill is consuming
the expected power, there is little concern about the class of
bodies using the supplied energy1.

Sensing load movement: the Impactmeter*

(patent pending)
In recognition of ore variability, many current SAG mills

are installed with variable speed drives, capable of

operating the mill through a wide and continuous range of
speed values. Mill speed and liner profile determine
grinding media trajectories, as shown in Figure 1. By
increasing mill speed or the steepness of liner attack angle,
grinding media may be inadvertently permitted to strike on
the mill liners instead of being directed to impact on the
mill load (see Figure 2), resulting in wasting large amounts
of energy, damaging and breaking mill liners (Figure 3 and
Figure 4) and promoting accelerated breakage and
degradation of grinding media (Figure 5). With
conventional instrumentation, the mill operator has no
sense of these effects, thus being unable to take appropriate
control measures. Even in the case where the operator is
conscious of this phenomenon, he has little or no feedback
from instrumentation to check the extent of the effect of his
control actions. These circumstances create a need for new
instrumentation to sense load movement effects.
The load movement inside of a SAG mill can be
monitored through the audible impacts of the falling stream
of ore and grinding media with themselves and with the
mill liners. The mill noise is a mixture of many different
sound sources. The main source is the large number of
small shocks between load components during tumbling of
the mill load, audible as a continuous stochastic noise. An
example of the waveform of this noise may be found in
Figure 6, obtained on a 36 diameter mill running at 8 rpm
and well filled with mineral. When the minerals hold-up in
the mill decreases, the mill noise intensity grows inversely,
as shown in Figure 7. This effect is the basis of the electric
ear patented by Hardinge to control the mineral feed to ball
mills. When the falling stream hits the liners in free fall,
high energy impacts appear superimposed over the common
mill background noise, as shown in Figure 8. Figure 9
shows the amplitude spectra associated with each sound
sample, in relative units. It is easily seen that spectra are not
effective tools to differentiate between high- and lowimpact operating conditions.

*Impactmeter patent pending in Australia, Canada, Chile, South Africa

and USA

PERFORMANCE ENHANCMENT TOOLS FOR GRINDING MILLS

103

Figure 1. Effect of mill speed and lifter profile on load trajectors

b
Figure 2. Schematic view of ball trajectories. Load falling onto itself (a) and on the liners (b)

Figure 3. Examples of liner damage under impact

104

PLATINUM ADDING VALUE

Figure 4. Broken liners

Figure 5. Excessive speed and low mineral hold-up cause extensive ball breakage

Figure 6. Low-noise, low-impact operating conditions as displayed by CoolEdit software and reflected as sound amplitude over time

By locating microphones near the mill on fixed positions

and applying appropriate algorithms online to the
microphone signals, impacts on mill liners can be isolated
from mill noise, then classified and counted to provide a
quantitative measure of their occurrence. The observed
distribution is partitioned into two ranges and the counts on
each range are defined as standard impacts and critical

impacts. Critical impacts comprise mainly direct impacts

to the liners, as these show higher acoustical levels. This
method of monitoring impacts, together with the
specialized hardware and software developed for use with
the acoustic sensors, has been referred to in this paper as
the Impactmeter2.

PERFORMANCE ENHANCMENT TOOLS FOR GRINDING MILLS

105

Figure 7. High-noise, associated with a low mineral hold-up condition as displayed by CoolEdit software and reflected as
sound amplitude over time

Figure 8. Impacts superimposed over mill noise as displayed by CoolEdit software and reflected as
sound amplitude over time

Figure 9. Samples of mill sound spectra, under low- and high- impact conditions

106

PLATINUM ADDING VALUE

Figure 10. Time evolution of sound spectra for the three sound samples of Figure 6, Figure 7 and Figure 8. Time units are (minute:seconds)

Speed (RPM)

9.0

8.5
8.0

Standard Impacts

800
600
400
200
0
-200
11:31:1

11:34:2

11:37:2

11:40:3

Time

Figure 11. Variation of standard impacts, count with mill speed

The Impactmeter signals provide the information needed

for mill control. By maintaining the critical impact rate in a
low to nil value, the best instantaneous load movement
pattern is obtained, controlled either by a human operator or
the automatic control system. The main variable used for
acting on impact rate is mill speed. It has a direct effect, is
easily controllable, and presents little interference with
other variables. Figure 11 shows the dependence of
standard impact rate with mill speed. There is a minimum
speed to observe impacts. Low mill filling, coarse feed and
low solids percentage may also increase the standard
impact rate, and even the critical impact rate in extreme
cases. The appropriate control action differs in these cases.
Low mill filling and solids percentage must be
compensated per se acting on feed rate and water flow,
respectively. The increase of standard impact rate is derived
mainly from the lack of cushioning material at the contact
points, which has an amplifying role on the whole mill
noise. Mill speed can be lowered during the transient time

to reduce the impact rate, but it may be a secondary

consideration. Coarse feed effect is derived both from lack
of fine material for cushioning and from changes in fall
trajectories due to the associated higher position of load
shoulder. This last effect may be compensated with a slight
decrease in mill speed.
The effects and benefits of online controlling of the load
falling trajectories may be exemplified with the following
data, extracted from a Chilean SAG plant. Operating data
comprising a whole week was analysed. The data was
classified by tonnage in quartiles, selecting the quartile of
highest tonnage for analysis. Figure 12 is a plot of the mill
speed and bearing pressure for each point of the high
tonnage data set. It may be seen that these points populate
over high speed and high bearing pressure, signalling a
good mill filling. All process points showing high critical
impact rate are shown on the same plot. Most points are
situated outside the field of high tonnage conditions. Only
at the zone of high speed and medium bearing pressure is

PERFORMANCE ENHANCMENT TOOLS FOR GRINDING MILLS

107

Figure 12. Inverse relationship between high tonnage and high critical impact

Figure 13. Impact level as a function of speed and bearing pressure

there a degree of correlation of both data sets. By observing

Figure 13, there are also low impact points in this zone, due
to third variables acting on impact rate. The main benefit of
operating under low impact conditions by using the
Impactmeter signal is self evident from this data: better
tonnage and lower specific energy consumption. Additional
benefits related to reduced steel consumption in balls and
liners, and the reduced risk of liner breakage are also
evident.

Sensing load quality: the SAG Analyzer

(patent pending)
A second operational problem is the uncertainty of the

108

composition of mill load. Data obtained from load cells or

bearing pressure simply provide information about the load
weight, but do not provide any information about load
volume or apparent density. The varying nature of mill
charges is such that there can be many combinations of
balls, mineral and water having the same weight but
comprising differing proportions of these components. This

SAG analyser patent pending in Australia, Canada, Chile, South Africa

and USA

PLATINUM ADDING VALUE

uncertainty may be very misleading for the operator. It is

possible that a load composed largely of balls may be
interpreted as a very hard ore condition, due to the lower
tonnage and higher power consumption observed. Also, in
spite of efforts to keep the ball level at a constant value, ball
breakage or changes in the mineral feed may produce
marked variations that may remain hidden to the operator.
For these reasons, there is an additional need to know the
ball level inside the mill.
The load level and composition by percentage volume of
total mill volume may be determined through the use of
mill noise (see Figure 14), process data and software based
on appropriate algorithms. This method of determining mill
load composition, together with the software developed, has
been referred to in this paper as the SAG Analyzer2. Other
authors have reported work in this field310.
The SAG Analyzer performs a two-phase workflow.
During the first phase, a microphone located on the mill
shell sends acoustic data to a main unit, where the toe and
shoulder positions are measured by appropriate algorithms.
This data is fed to models, which give a close estimation of
the load volume. During the second phase, the ball level is
estimated from the load volume just obtained and data
extracted from the mill control system. The underlying

principle for this estimation is the analysis of the power

consumption patterns of the mill with a detailed power
model calibrated with plant data. The SAG Analyzer seeks
the ball level giving the best fit to a set of current power
data. Making use of these two values, the instrument also
calculates the instantaneous load density. This information
together with specific process variables extracted from the
plant control system, is used to generate, online, a screen
display showing total volumetric mill filling percentage,
ball volumetric filling percentage, ore volumetric filling
percentage, total charge apparent density, angular position
of charge toe, and angular position of charge shoulder. Mill
speed in revolutions per minute, bearing pressure, tonnage
throughput and power draw are also reflected on the SAG
Analyzer display monitor, as shown on Figure 16.
The main benefit of this new technology is the gain in
plant stability. The reduction of uncertainty about ball level
does allow better control actions by a better interpretation
of the mill behaviour, and assists in more easily identifying
the source of concerns. Knowledge of the load volume
allows a better management of mill filling, maintaining a
higher mean hold-up and facilitates the processing of higher
tonnages without the risk of overfilling the mill. An
advantage of this specific technology is that it is based on

Figure 14. Sound trace used to find the charge position during a revolution of the mill. The toe is near the highest amplitude points

Position
reference
plate

wireless
acoustic sensor
module

Inductive
sensor

receiver

main unit

Figure 15. Layout of the microphone and position sensor in the mill

PERFORMANCE ENHANCMENT TOOLS FOR GRINDING MILLS

109

Figure 16. SAG analyser** Monitor display

direct measurement of load toe and shoulder positions,

thought to be a reliable source of data to support the system
operation and quality of information. At the time of
preparing this paper, the technology applied to the SAG
Analyzer has only been recently installed and no long-term
quantitative studies are readily available to fully evaluate its
effect on plant performance.

Conclusion
The use of acoustics in conjunction with appropriate
algorithm based software can be used effectively to monitor
load movement within a SAG mill. The information
provided from these methods can be employed to control
mill behaviour and enhance mill performance.

Acknowledgement
The authors acknowledge the important contributions
provided from Federico Santa Maria Technical University,
the University of Santiago de Chile and the Chilean Council
for Scientific and Technological Research through
FONDEF, in the development of this technology.

References
1. VALDERRAMA, W. and L MAGNE, L. Notes on
Operation, Maintenance and Control of
Semiautogenous Grinding Plants for short training
courses (unpublished work in Spanish).
2. SAMS, C., NARANJO, G., and KEMERER, J. New
Performance Enhancement Technologies for the
Milling Industry, Workshop on SAG Grinding,
SAG03, Via del Mar, Chile, October 2003. CDROM.
3. JKtech Technology ToolsDynagrind, November
2003. CD-Rom.
4. CARTES, F., RIVAS, G., MUJICA, F., and

110

SCHWARZE, K. On-line estimation of the inclination

of the charge in a SAG mill. Proceedings of COPPER
95COBRE 95 International Conference vol. II
Mineral Processing and Environment. A. Casali,
G.S.Dobby, M.Molina and W.J.Thoburn (eds.). The
Metallurgical Society of CIM, 1995. pp. 225232.
5. CAMPBELL, J., SPENCER S., SUTHERLAND, D.,
ROWLANDS, T., WELLER, K., CLEARY, P., and
HINDE, A. SAG mill monitoring using surface
vibrations, Proceedings International autogenous and
semiautogenous grinding technology 2001, Sep. 30
Oct. 3, 2001, Barratt et al. (eds.), CIM, vol. IV,
pp. 373-385.
6. PAX, R. Non-Contact acoustic measurement of inmill variables of SAG mills, Proceedings
International autogenous and semiautogenous
grinding technology 2001, Sep. 30 Oct. 3, 2001,
Barratt et al. (ed.), Published CIM, vol. II,
pp. 387393.
7. SPENCER, S.J., CAMPBELL, J.J., WELLER, K.R.,
and LIU, Y. 2000 Monitoring of SAG Mill
Performance Using Acoustic Emissions. Proceedings
of the XXI International Mineral Processing
Congress, Rome, July 2000, A413, A420.
8. WATSON, J.L. An Analysis of Mill Grinding Noise,
Powder Technology, vol. 41, no. 1, 1985. pp. 8389.
9. ZENG, Y. and FORSSBERG, E. Monitoring
Grinding Parameters by Vibration Signal
MeasurementA Primary Application, Minerals
Engineering, vol. 7, no. 4, 1994. pp. 495501.
10. KALKERT, P. Mill filling level sensor: an electronic
ear for all situations, Zement, Kalk Gips INT 57,
2004, pp. 5559.

PLATINUM ADDING VALUE